regulators
regulators
I have some thoughts on this issues people are having here, and the increasing complexity of peoples designs versus other folks DIY skills and willingness to deal with custom PCBs, SMT parts, etc.
I recently had a discussion of the merits of CAT4101 driver on another thread, that can be summed up to "high efficiency and lower heat dissipation because it has lower dropout voltage". The thing is that there are alot of great regulators out there, better than this one by a good bit in fact, but they are limited on input voltage.
Now I'm not totally sure this will work, but I strongly think so. We can run the regulator on the low side of the circuit, having the LEDs feed power into the Vin line. By doing this, the highest voltage seen by a 6-LED string on a 24V supply is about 3 volts, meaning some really great regulators become available to us. This also open up the potential to drive more LEDs from the same driver... For instance 25 LEDs on a 90V supply is only 2.5V at the regulator, so you can use the exact same driver. (Of course this only works for LDO type regulators). I saw someone trying to do this on another thread, and would be super-simple this way.
Can anyone think of a reason this wont work? FYI I just tried it at 24V and it works perfectly... checked voltages at all pins vs ground with the multi-meter too and everything is about right.
Second, I'm not sure everyone is doing this but you should ALWAYS ALWAYS ALWAYS run these regulators in current-controlled mode. This eliminates most of the problems are seeing with variation in LEDs voltages. LEDs are not voltage-based devices, they are current-based.
Hi!
You are absolutely right. If you use an open-drain current sink (there are many of these, and this is the name you should search for to find similar parts), you can connect up an arbitrary number of LEDs in a chain.
The limiting factor on these is mostly in the heat dissipation. The regulator at the bottom is usually linearly regulating whatever voltage is left at the bottom of the chain down to ground at a fairly high current. For example, the 2.5V you suggest at 700mA is producing 1.75W of heat, and that's a fairly benign example.
The biggest problem you will find is that LEDs have fairly wide forward voltage tolerances. So while they might be typically at 3.2V, the *actual* Vf at 700mA might vary from 2.6 to 4.0V! Now, statistically if you had a random sampling and the parts were distributed with a normal function around 3.2V, adding many LEDs would result in a tighter average. However, for any parts you purchase you are likely to see a strong correlation between them because they will be from the same manufacturing line at the same time. So if one LED is 2.9V instead of 3.2V, they are *all* likely to be.
So instead of those errors balancing out, they will compound each other. With a 25 LED chain, that fairly small .3V error in the datasheet means a variation in the total voltage drop of 7.5V. So suddenly, the 2.5V you calculated you'd be dropping at the bottom is *10V* and now you're dissipating 7W from a relatively small SOIC chip.
Even worse, you have to design for the worst case in the *other* direction. If you don't have at least a volt or two at the bottom, you have no regulation capacity, so you have to assume the worst, in which case you're assuming a 4V drop in the design phase, but only a 2.6V drop is actually present! This 1.4V less Vf per LED means you're now dropping *35V* at the bottom, and trying to dissipate almost 25W! That will melt just about any standard chip.
So, the only way to design a system that does this reliably in practice if you are using one of these open-drain linear regulators is to do it completely custom for each system. That way you have measured instead of estimated values for the total Vf. In other words, if you happened to have 35V across that last LED to the regulator, you would just throw on another 12 LEDs, or drop the input voltage.
Another way to help with this is to put in a power resistor at the top of the LED chain to drop the current somewhere other than the main regulator. That spreads out the heat generation beyond the more fragile silicon.
Personally, I highly recommend switching regulators such as the LM3404 and LM3406 for these sorts of high power applications. They will not have any of these problems, but they do have those voltage limits that you won't see otherwise (because they are controlling a power transistor at the top of the chain).
Best,
Brian
) If something just short in a usual way the driver will shutdown automatically and it is no hazard to anyone or anything.
